Help with a project called "The Ultimate Last Layer Algorithms Collection"

[GRVigo]

Recently I have been dedicated to improving my solver. I have rewritten it from scratch, and now it is faster, along with other improvements, but I still need to redesign the user interface, which, at the pace I’m going, will take me a few more months before I can make it available to everyone. I also set out to improve the collections of algorithms I use, and taking advantage of the solver’s code, I adapted it to generate algorithms, and then… I was possessed by the hoarding syndrome of last layer algorithms: I am generating all algorithms of 15 moves or less, as well as algorithms with more moves with some simplifications. The number of algorithms I have collected is overwhelming: I have generated more than two million unique algorithms, and considering that the same algorithm can be written in thousands of different ways (modifying its notation by introducing twists and/or substituting moves with equivalent ones), it turns out that I literally have billions of last layer algorithms to choose from!

To make all these algorithms useful, I have programmed some tools to classify them and assign them a score. The selection of the best ones is based on three parameters: their length or metric, the type of moves they consist of, and whether they contain certain sequences (triggers). Thus, I have generated some PDF documents of algorithm collections that I would like to make available to everyone. Right now, I have preliminary versions of PLL, OLL, COLL, ZBLL, AntiZBLL, and 1LLL with the best algorithms according to these criteria.

Here is an example:


If any forum user is interested, please contact me privately, and I will provide a link to download these preliminary collections in PDF. Before releasing them to everyone, I would like to get some opinions on:

- whether the way of presenting the algorithm sheets is well understood

- are the case names appropriate?

- whether the selection of algorithms is adequate, could you give me examples of algorithms that do not appear and, in your opinion, should appear?

- I could generate collections with many more algorithms, would that be of interest?

- what could be added/changed to improve this collection of algorithms?

I have other ideas in mind. For example, by modifying the scoring parameters, the best algorithms for one-handed mode or for left-handed people could be obtained. I could also prepare an application to obtain personalized algorithm collections for each person, either by entering the scoring parameters to the user’s liking or by specifying some algorithms as a reference and having the application propose similar algorithms for different cases. If you have other possible applications or ideas, or any comments, I would be happy to hear them.

Thank you for your attention

 

[Filipe Teixeira]

inventor of CFOP: let's keep it simple by using only 57 orientations and 21 permutations so the user can focus solely on crappy algs and intuition

this guy:

it turns out that I literally have billions of last layer algorithms to choose from!

 

[abunickabhi]

Recently I have been dedicated to improving my solver. I have rewritten it from scratch, and now it is faster, along with other improvements, but I still need to redesign the user interface, which, at the pace I’m going, will take me a few more months before I can make it available to everyone. I also set out to improve the collections of algorithms I use, and taking advantage of the solver’s code, I adapted it to generate algorithms, and then… I was possessed by the hoarding syndrome of last layer algorithms: I am generating all algorithms of 15 moves or less, as well as algorithms with more moves with some simplifications. The number of algorithms I have collected is overwhelming: I have generated more than two million unique algorithms, and considering that the same algorithm can be written in thousands of different ways (modifying its notation by introducing twists and/or substituting moves with equivalent ones), it turns out that I literally have billions of last layer algorithms to choose from!

To make all these algorithms useful, I have programmed some tools to classify them and assign them a score. The selection of the best ones is based on three parameters: their length or metric, the type of moves they consist of, and whether they contain certain sequences (triggers). Thus, I have generated some PDF documents of algorithm collections that I would like to make available to everyone. Right now, I have preliminary versions of PLL, OLL, COLL, ZBLL, AntiZBLL, and 1LLL with the best algorithms according to these criteria.

Here is an example:
View attachment 34069

If any forum user is interested, please contact me privately, and I will provide a link to download these preliminary collections in PDF. Before releasing them to everyone, I would like to get some opinions on:

- whether the way of presenting the algorithm sheets is well understood

- are the case names appropriate?

- whether the selection of algorithms is adequate, could you give me examples of algorithms that do not appear and, in your opinion, should appear?

- I could generate collections with many more algorithms, would that be of interest?

- what could be added/changed to improve this collection of algorithms?

I have other ideas in mind. For example, by modifying the scoring parameters, the best algorithms for one-handed mode or for left-handed people could be obtained. I could also prepare an application to obtain personalized algorithm collections for each person, either by entering the scoring parameters to the user’s liking or by specifying some algorithms as a reference and having the application propose similar algorithms for different cases. If you have other possible applications or ideas, or any comments, I would be happy to hear them.

Thank you for your attention

PLL, OLL, COLL, ZBLL, AntiZBLL, and 1LLL is a nice first level classification.

I am exhausted by 5-style development, so I do not have the bandwidth to contribute to your endeavour.

Is there a better way to get help for you other than sharing PDFs privately to interested cubers?

 

[GRVigo]

PLL, OLL, COLL, ZBLL, AntiZBLL, and 1LLL is a nice first level classification.

I am exhausted by 5-style development, so I do not have the bandwidth to contribute to your endeavour.

Is there a better way to get help for you other than sharing PDFs privately to interested cubers?

Trying to master 5-style seems like an impossible task to me; I’m already happy knowing the Old Pochman method


I uploaded the algorithms collections to Mega, you can download them here:

TULLAC-V09b-Preliminary.zip

There is a single ZIP file of 82,9 MB

I would appreciate it if you could take a look at it, in case you have any ideas to improve the presentation or spot any errors I might have missed. Since I’ve put in the effort to prepare it, I would like it to be as practical as possible.

Thank you for your attention

 

[OreKehStrah]

Trying to master 5-style seems like an impossible task to me; I’m already happy knowing the Old Pochman method


I uploaded the algorithms collections to Mega, you can download them here:

TULLAC-V09b-Preliminary.zip

There is a single ZIP file of 82,9 MB

I would appreciate it if you could take a look at it, in case you have any ideas to improve the presentation or spot any errors I might have missed. Since I’ve put in the effort to prepare it, I would like it to be as practical as possible.

Thank you for your attention

I just took a glance through and I think it looks great visually!

I do question the case ordering though. The order the OLL sets are placed seemed to be a bit random to me. I’ll take a further look when I get off work to see if I missed anything or have more specific suggestions for how it could be arranged, though I do have some ideas already.

 

[GRVigo]

I do question the case ordering though. The order the OLL sets are placed seemed to be a bit random to me. I’ll take a further look when I get off work to see if I missed anything or have more specific suggestions for how it could be arranged, though I do have some ideas already.

The algorithms are sorted by "quality." The best of each case has three stars and those with a score up to 5% worse too. Up to 10% for those with two stars and one star for those with up to 20%. I'd should explain this before, sorry.

Thank you for your kind comments.

 

[OreKehStrah]

The algorithms are sorted by "quality." The best of each case has three stars and those with a score up to 5% worse too. Up to 10% for those with two stars and one star for those with up to 20%. I'd should explain this before, sorry.

Thank you for your kind comments.

Oh I meant the order of the cases, not the algorithms. I could already tell there was some sort of ranking metric. I do think it needs to be tweaked because there were cases with some questionable algs at the top just scrolling through.

In the 1LLL current ordering it goes from PLL into H set ZBLL which I find to be a strange set to put right after.

 

[GRVigo]

Oh I meant the order of the cases, not the algorithms. I could already tell there was some sort of ranking metric. I do think it needs to be tweaked because there were cases with some questionable algs at the top just scrolling through.

The scoring of algorithms is somewhat subjective. I have in mind to prepare an application to be able to adjust the scoring parameters and obtain personalized rankings.

In the 1LLL current ordering it goes from PLL into H set ZBLL which I find to be a strange set to put right after.

Cases of 1LLL are usually in clusters of 72. But ZBLL H are only 40, which added to the 22 of PLL (if we count the Skip case) we almost have a complete group, that's why I grouped them that way.

 

[OreKehStrah]

The scoring of algorithms is somewhat subjective. I have in mind to prepare an application to be able to adjust the scoring parameters and obtain personalized rankings.


Cases of 1LLL are usually in clusters of 72. But ZBLL H are only 40, which added to the 22 of PLL (if we count the Skip case) we almost have a complete group, that's why I grouped them that way.

I figured that was the case. I just don’t think it’s a good way to present them since some people might interpret that H ZB should be next after PLL

 

[abunickabhi]

Trying to master 5-style seems like an impossible task to me; I’m already happy knowing the Old Pochman method


I uploaded the algorithms collections to Mega, you can download them here:

TULLAC-V09b-Preliminary.zip

There is a single ZIP file of 82,9 MB

I would appreciate it if you could take a look at it, in case you have any ideas to improve the presentation or spot any errors I might have missed. Since I’ve put in the effort to prepare it, I would like it to be as practical as possible.

Thank you for your attention

Thanks for sharing the zip file. The algs look good visually, let me take a deeper look.